An Israeli retail chain was making a lot of noise with aggressive "buy two, get one" deals — and quietly working for nothing, because no one costed the promotions in real time. We put a guardrail between a loss-making deal and the till.
The deals pulled huge footfall into the branches. By the end of the month, the gross margin had been ground into the floor — the chain was working hard to sell at a loss.
Three things drove it. Promotions that burned money: deals went out to chase sales, but nobody calculated the product's real, fully loaded cost — logistics, shipping, operations — so stock was sold below what it cost to move. The truth arrived too late: it took finance nearly two weeks of manual work to pull data from the tills and the accounts and see whether a promotion had worked — long after it had ended and the damage was done. And manual pricing at the till: every price-list change meant hand-keying thousands of product codes into old till systems, a sea of errors and shelf-versus-checkout gaps.
Real cost, computed live, with a hard line no promotion can cross.
We built a tool that pulls live data from the point-of-sale and the company's accounting system and computes the real net profit on every single item — every associated cost included, not just the sticker price minus the buy price.
An iron rule in the system's logic: if a promotion's pricing erodes gross margin below a 12% floor, the system blocks it from going live automatically — and fires an alert to the trade manager at the same moment.
Approved promotion prices flow directly from the engine to the tills in the branches at the press of a button — no hand-keying, no exhausting re-entry, no gap between the engine and the shelf.
The fortnight of manual reconciliation to find out whether a deal worked collapsed to a live read. Trade sees the true margin of a promotion as it runs, not after it's over.
Because prices move by integration rather than by typing, the mismatches between the price on the shelf and the price at the checkout — and the errors that came with manual entry — simply stopped.
The system does the maths and holds the line; a person still sets the strategy. When a deal is blocked, the manager sees exactly why and can rework it — informed, in the moment, not weeks later.
Pricing moved from something discovered after the fact to something managed in real time.
An immediate improvement of around 4% in the chain's gross margin, driven mostly by catching and stopping loss-making campaigns and promotions in real time — before they ever reached a till.
Pricing errors and shelf-versus-checkout gaps were eliminated entirely, and the trade team moved from managing profitability in hindsight — working out last month's losses — to steering it accurately, day to day.
What makes it durable is the guardrail: a hard margin floor the system enforces every time, on every item, without a tired human having to remember to check. The AI does the costing and holds the line; the people set the strategy above it. That is the discipline that turns a clever pricing tool into one a business can actually trust in production — the same principle behind our data-readiness work.
Reef TRH